Peter Eisentraut <peter.eisentraut@enterprisedb.com> writes:
> On 19.04.22 16:21, Tom Lane wrote:
>> * In the other direction, decide that the PG_BINARY_X macros are
>> offering no benefit at all and just rip 'em out, writing "rb" and
>> so on in their place. POSIX specifies that the character "b" has
>> no effect on Unix-oid systems, and it has said that for thirty years
>> now, so we do not really need the platform dependency that presently
>> exists in the macro definitions. The presence or absence of "b"
>> would serve fine as an indicator of intent, and there would be one
>> less PG-specific coding convention to remember.
> I can only imagine that there must have been some Unix systems that did
> not understand the "binary" APIs required for Windows. (For example,
> neither the Linux nor the macOS open(2) man page mentions O_BINARY.)
> Otherwise, these macros don't make any sense, because then you could
> just write the thing directly on all platforms.
PG_BINARY is useful for open(). It's the PG_BINARY_R/W/A macros for
fopen() that are redundant per POSIX. Possibly someone generalized
inappropriately; or maybe long ago we supported some platform that
rejected the "b" option?
regards, tom lane
